1. Nonlinear systems analytic techniques were used to characterize transformational properties of the network of neurons activated by perforant path input to the rabbit hippocampus. Trains of 4,064 impulses with randomly varying interimpulse intervals were used to stimulate perforant path fibers, and amplitudes of evoked dentate granule cell population spikes were measured. Interimpulse intervals of the random stimulus train were determined by a Poisson distribution with a mean interimpulse interval of 500 ms, and with intervals ranging from 1 to 5,000 ms. The response of dentate granule cells to this stimulation was assumed to reflect activity in the larger hippocampal network, because other subpopulations of neurons activated monosynaptically and polysynaptically within the hippocampal formation contribute to granule cell excitability through multiple feedforward and feedback pathways. System properties were characterized both for halothane anesthetized and chronically implanted, unanesthetized preparations. 2. Second-order kernel analysis showed that population spike amplitude was highly dependent on interimpulse interval. When population spikes of all latencies were included in the same analysis, stimulation impulses produced near-total suppression of spike amplitude when they were preceded 10-20 ms by another impulse in the train. Spike suppression extended to approximately 50 ms and was inversely related to length of the interimpulse interval. Suppression of granule cell response to intervals within the range of 10-50 ms was not influenced by halothane anesthesia. 3. Interstimulus intervals greater than approximately 50 ms resulted in a facilitation of population spike amplitude, with maximum facilitation occurring in response to intervals of 90-100 ms. The magnitude of maximum facilitation was significantly greater for anesthetized (129%) than for unanesthetized (74%) preparations. The range of intervals resulting in facilitation for unanesthetized animals could extend to 1,000-1,100 ms (average range, 61-714 ms). This was much greater than observed for population spikes recorded from anesthetized animals (50-364 ms), which exhibited suppression in response to intervals of approximately 300-700 ms. 4. Further analysis revealed that the nature of nonlinearities in population spike amplitude may depend on spike latency. For example, population spikes of "short" latency (3-4 or 4-5 ms, depending on the animal) exhibited only facilitation in response to interstimulus intervals of 1-4 ms.(ABSTRACT TRUNCATED AT 400 WORDS)